Skip to content

[SPARK-29244][Core] Prevent freed page in BytesToBytesMap free again #25953

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
viirya wants to merge 4 commits into from
Closed

[SPARK-29244][Core] Prevent freed page in BytesToBytesMap free again #25953

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
dbd3435
Prevent freed page free again.
viirya Sep 27, 2019
9d5c2d1
Simplify fix.
viirya Sep 28, 2019
742fb45
Add test.
viirya Sep 29, 2019
441aa2b
Address comments.
viirya Oct 1, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions core/src/main/java/org/apache/spark/unsafe/map/BytesToBytesMap.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -886,6 +886,7 @@ public void reset() {
numKeys = 0;
numValues = 0;
freeArray(longArray);
longArray = null;
while (dataPages.size() > 0) {
MemoryBlock dataPage = dataPages.removeLast();
freePage(dataPage);
Expand Down
19 changes: 19 additions & 0 deletions core/src/test/java/org/apache/spark/unsafe/map/AbstractBytesToBytesMapSuite.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,6 +34,7 @@
import org.apache.spark.SparkConf;
import org.apache.spark.executor.ShuffleWriteMetrics;
import org.apache.spark.memory.MemoryMode;
import org.apache.spark.memory.SparkOutOfMemoryError;
import org.apache.spark.memory.TestMemoryConsumer;
import org.apache.spark.memory.TaskMemoryManager;
import org.apache.spark.memory.TestMemoryManager;
Expand Down Expand Up @@ -726,4 +727,22 @@ public void avoidDeadlock() throws InterruptedException {
}
}

@Test
public void freeAfterFailedReset() {
Copy link
Contributor

@JoshRosen JoshRosen Oct 1, 2019

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Mind adding a one-line comment in the body of the test noting that this is a regression test for SPARK-29244?

Copy link
Member Author

@viirya viirya Oct 1, 2019

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ok. Added.

// SPARK-29244: BytesToBytesMap.free after a OOM reset operation should not cause failure.
memoryManager.limit(5000);
BytesToBytesMap map =
new BytesToBytesMap(taskMemoryManager, blockManager, serializerManager, 256, 0.5, 4000);
// Force OOM on next memory allocation.
memoryManager.markExecutionAsOutOfMemoryOnce();
try {
map.reset();
Assert.fail("Expected SparkOutOfMemoryError to be thrown");
} catch (SparkOutOfMemoryError e) {
// Expected exception; do nothing.
} finally {
map.free();
}
Copy link
Contributor

@JoshRosen JoshRosen Oct 1, 2019 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Even though this test suite fails without the longArray = null change, it turns out that another (wrong!) way to make this test pass would be to simply remove the free(longArray) call and leak memory.

Unfortunately, though, that doesn't cause the test to fail! It turns out that the

  @After
  public void tearDown() {
    Utils.deleteRecursively(tempDir);
    tempDir = null;

    if (taskMemoryManager != null) {
      Assert.assertEquals(0L, taskMemoryManager.cleanUpAllAllocatedMemory());
      long leakedMemory = taskMemoryManager.getMemoryConsumptionForThisTask();
      taskMemoryManager = null;
      Assert.assertEquals(0L, leakedMemory);
    }
  }

doesn't work as expected because TestMemoryManager doesn't properly implement per-taskAttemptId memory-consumption tracking.

I've opened #25985 to fix that problem. That doesn't need to block merging of this patch, however.

}

}